How we keep this case study neutral
Adopted October 31, 2026. Linked from every page of the case study; changes to this standard are dated and listed in the corrections log.
This case study describes the 2026 midterm elections: who ran, who won, how places voted compared with before, and what those places are like. It does not tell anyone how to vote, what to think of a party or candidate, or what the results mean for the country. These are the rules we follow to make sure of that. Every page and every sentence is checked against them before it is published.
1. We describe; we do not judge
- We report what happened and what the data shows: votes, shares, margins, changes, and the characteristics of places.
- We do not predict elections, rate races, endorse anyone, or say whether a result is good or bad.
- We avoid words that take sides or carry a verdict. We write "won by 2.4 points", not "narrowly survived"; "the share fell by 6 points", not "collapsed". A list of words we do not use is kept with this standard and checked automatically.
- We do not explain results with motives, campaigns or events. Where a finding invites an explanation, we say what the data does and does not show, and leave the explanation to the reader.
2. Every party and candidate is shown the same way
- Every candidate on the ballot is listed, including third-party, independent and write-in candidates with a reported count.
- Candidates are listed by votes received, and before results by the order the state's ballot uses.
- Party colours are used the same way for every party, with the same strength of colour for the same size of result. Colour scales are symmetrical: a 10-point lead looks the same whichever party has it.
- The same rounding, the same scales and the same chart types are used for every party and every place.
- Maps are offered in two views: the normal map and an equal-size view in which every district or state is the same size, so large, thinly populated areas do not look more important than they are.
3. We never call a race
- We show each race's status as the official count gives it: counting, unofficial result, runoff, recount, certified.
- Where a news organisation has projected a winner, we say who made the projection and when. We make no projections of our own.
- Unofficial figures are marked unofficial, with the time we read them. Certified figures replace them when each state certifies, and every change is logged.
4. Places, not people
- Results by county or district are compared with what those places are like: income, age, education, race and ethnicity, commuting, housing and more. These comparisons describe places. They do not say how any person or group voted, and every such view says so.
- We do not estimate how groups of voters voted unless the estimate comes from a published, credited source, and we name that source.
- A pattern is never presented as a cause. Each finding states how strong the pattern is, where it holds and where it does not, and what other explanations the data cannot rule out.
5. Every figure has a source and an honest margin
- Every figure names its source. Census figures are estimates from a survey and are shown with their margin of error and a reliability rating; figures too uncertain to rely on are marked.
- Where we build a figure ourselves (for example, Census figures for districts whose lines changed in 2026), we say so and say how.
- Where sources disagree or contain errors, we say what we found and what we did about it.
6. Polls
- We use only polls whose publishers allow reuse, and we credit each pollster.
- We compare final polls with results to show how accurate they were, for every party alike. We build no forecast of our own.
7. Corrections and openness
- Mistakes are corrected as soon as we find them, and every correction is listed, with its date, in a public corrections log.
- The data behind every view can be downloaded, and the methods are published, so anyone can check our work.
- Everything we publish may be reused with credit (Creative Commons Attribution 4.0). Source data keeps its own terms.
8. How the standard is applied
- Before the first edition, two readers with different political views and the founder of Spotbookr review every screen and every sentence against this standard.
- After that, every update gets a shorter check against the standard before it is published, and the "What's new" log records it.
- Anyone who believes something in the case study breaks this standard can tell us, and we will answer publicly in the corrections log.
Spotbookr Inc. builds data and mapping software. This case study is free and is not funded by, or produced for, any party, candidate, campaign or political organisation.